The Spatial Pattern of Urban Land Prices
Determination in Indonesia
The Case of Surabaya, East Java
A Research Paper presented by:
Puspa Anggraini
(Indonesia)
in partial fulfilment of the requirements for obtaining the degree of
MASTER OF ARTS IN DEVELOPMENT STUDIES
Major:
Economics of Development
(ECD)
Members of the Examining Committee:
John Cameron
Mansoob Murshed
The Hague, The Netherlands
August 2013
Contents
List of Tables
iii
List of Figures
iii
List of Maps
iv
List of Appendices
iv
List of Acronyms
iv
Aknowledgement
vi
Abstract
vii
Chapter 1 Introduction
1
1.1
Background
1
1.2
Problem statement
2
1.3
Research question
3
1.3.1 Main research question
3
1.3.2 Sub question
3
Structure of the paper
4
1.4
Chapter 2 Literature Review
4
2.1
Basic theory of land value
4
2.2
Urban land prices determinants and its analysis
5
2.3
Type of urban structure
10
Chapter 3 Background of Study Area
14
Chapter 4 Data and Methodology
17
4.1
Data
17
4.2
Methodology
27
4.3
Data description
29
Chapter 5 Results and Analysis
33
5.1
Descriptive Analysis
33
5.1.1 CBD and land prices
33
5.1.2 Non distance variables
36
5.1.3 Distance variables
37
5.1.4 Central and non-central analysis
41
Regression results and analysis
42
5.2.1 Regression of land prices and CBD
42
5.2
i
5.2.2 Pooled regression
43
5.2.3 Principal Component Analysis
44
5.2.4 Regression result using principal components
46
Chapter 6 Conclusion and Recommendation
49
6.1
Conclusion
49
6.2
Recommendation
51
References
55
ii
List of Tables
Table 4.1 Dependent and independent variables
28,29
Table 4.2 Summary statistics of dependent and independent variables
30
Table 4.3 Distribution of land prices
31
Table 4.4 Summary statistic of land prices in central and non-central area
32
Table 5.1 Average land prices based on the distance from CBD
33
Table 5.2 Summary statistic of land prices’ ring
34
Table 5.3 Detail information of outliers
35
Table 5.4 Average land prices based on land use and zoning classification
36
Table 5.5 Average land prices related to road class and land tenure
37
Table 5.6 Average distance from variables to CBD
38
Table 5.7 Result of Mann-Whitney test
41
Table 5.8 Result of Kruskall-Wallis test
41
Table 5.9 Regression result of land prices and CBD
42
Table 5.10 Pooled regression result in land prices determination
43,44
Table 5.11 Eigenvalues and the proportion of variation explained by principal
components
44,45
Table 5.12 Principle component coefficients
45
Table 5.13 Principle component coefficients with omitted variables
46
Table 5.14 Regression result using principle component
47
List of Figures
Figure 2.1 Concentric urban zone
Figure 2.2 The radial urban zone
Figure 2.3 Hoyt sector zone
Figure 2.4 Multi-nuclei urban model
Figure 2.5 Sector zone model
Figure 5.1 Land prices that fall with distance
Figure 5.2 Land prices that rise with distance
iii
10
11
12
12
13
38,39
40
List of Maps
Map 3.1 Great Metropolitan Surabaya Area
Map 3.2 Administrative map of Surabaya
Map 4.1 Land valuation map
Map 4.2 Land utilization of Surabaya, 2009
Map 4.3 Zoning map of Surabaya
Map 4.4 Location of CBD and Sub CBDs
Map 4.5 Location of traditional market
Map 4.6 Location of public transportation and infrastructure
Map 4.7 Location of industrial area
Map 4.8 Location of final disposal site
Map 4.9 Location of openness area
Map 4.10 Distribution of land prices
Map 4.11 Ring of land prices distribution to CBD
14
15
18
19
20
22
23
24
25
26
27
31
34
List of Appendices
Appendix 1 Satellite images of outliers
Appendix 2 VIF result
52,53
54
iv
List of Acronyms
BPN
Badan Pertanahan Nasional
CBD
Central Business District
EDA
Explorotary Data Analysis
GDRP
Gross Domestic Regional Product
Gerbangkertosusilo
Gresik, Bangkalan, Mojokerto, Surabaya, Sidoarjo,
Lamongan
GIS
Geography Information System
HGB
Hak Guna Bangunan
HM
Hak Milik
HP
Hak Pakai
OLS
Ordinary Least Square
PCA
Principal Component Analysis
SAR
Spatial Auto Regressive
TPA
Tempat Pembuangan Akhir
VIF
Variance Inflation Factor
v
Acknowledgement
For almost two years, I have been experienced many valuable things in
academic and non-academic. Taking the opportunity to study in master
programme in ISS give me better understanding about the diversity around the
world.
By finishing my research paper, it means that my time as student is
done. Hence, I would take this opportunity to write my gratitude to all the
person that help me through joyful and difficult time. First, I would like to
thank to my parents, my brother and sister for all of your support. Then, my
deepest gratitude is to my friends from ECD group. In particular, thank you to
my entire Indonesian student that has been my new family for the last two
years.
Lastly, I would like to thank to all of my lecturers for the transfer of
valuable knowledge. In addition, my gratitude is to my supervisor for the
guidance, ideas and support, and my second reader for the valuable comment
during my research paper period.
vi
Abstract
This paper aims to examine the determinants of land prices linked to the
spatial pattern of urban area in one of the largest city in Indonesia that is
Surabaya. By using market-prices approach, this research incorporates 14
spatial variables and 4 non-spatial variables. The findings indicate that
proximity to city centre is still significant in determining land prices in
Surabaya. It fits with empirical findings from previous study such as Boston
Dallas and Jakarta. Nevertheless, the R-squared of land prices and CBD is
quite small indicating the urban structure in this city likely follow the
polycentric city or multi-nuclei city. It also confirms that there is a
decentralisation in urban economic activities in Surabaya.
Due to multicollinearity problem, regression in this research using four
principal components. The result show that there are four significant
independent variables which are first principle component (central area),
fourth principle component (outer area), zoning for industry and collector
road. Beside outer area, all the results can be supported by the descriptive
analysis and previous studies. It is likely because the second component the
consist of variables that tend to fall and rise with distance. In addition, land
prices in central area tend to decline with distance. In term of zoning, industrial
area has negative sign implying that this zoning has lower prices compared to
trading/service zone. Coefficient for collector road has positive sign indicating
that land plot in front of this road is appreciated higher values than located
near local road.
Relevance to Development Studies
Urban structure and land prices have a link with development in one city. For
instance, spatial pattern such as monocentric city that tend to accommodate
city centre will likely cause the uneven development between central and
periphery region. As a result, perhaps there are not equally distributed the
provision of public services around the city. Some cities based on previous
study such as Chicago, Istanbul and Berlin adopted that pattern in the early of
20th-century which can be seen from the steeper land prices gradient. However,
those metropolitan cities nowadays has transformed to polycentric city and
experienced the declining of importance of central business district (CBD).
In addition, some study indicated that land prices in city centre still
tend to relatively higher than in suburb area particularly in metropolitan area.
Higher land prices are the burden for government to accelerate development in
one city. It would be a restriction to build infrastructure such as road, parks
and other public facilities in order to support economic activities. It is also
disadvantage for citizen to get affordable housing. Related the note above, it is
worthy to examine the pattern of urban structure in Surabaya as the second
metropolitan city in Indonesia after Jakarta related to the dynamic of spatial
pattern of land prices.
vii
Keywords
Land prices determination, Spatial pattern of land prices, Surabaya, Indonesia.
viii
Chapter 1
Introduction
1.1 Background
Land according to Koomen and Buurman (2002) has unique characteristics
compared to other economic goods. The supply of land as a simple area is
fixed, every parcel of land has a fixed location, and the use of a parcel of land
affects the use and value of surrounding parcels. Similarly, Fujita (1989) stated
that land is a complex resource with dual characteristics. First, land is a
commodity in the usual economic sense, capable and being used in many ways.
However, second, unlike other commodities, land is immobile. Therefore, each
piece of land is associated with a unique location in geographical sense.
According to Harvey (1992) in his book Urban Land Economics,
underdeveloped land or ‘pure’ land refers solely to natural resources in space.
Thus, pure land can be regarded as being fixed in supply and therefore price
inelastic. But, land prices, like the prices of other goods, are influenced by
demand. Since land as a whole is a fixed supply provided by the nature, the
earnings of ‘pure’ land are determined solely by demand.
Related to development, differing land use particularly in a city are
closely related to with other factor. Land use in an urban area is driven by
many factors such as market, social or political factors. Harvey (1992) wrote
that a city begins for many reasons but economic forces are likely to be the
major factor influencing the urban growth. That expansion changes the pattern
of urban land use. In line with that, Barlowe (1978) quoted in Ai (2005)
explains that the allocation and use of land resources are affected by the
physical and biological, economic and institutional factors. Furthermore, the
dynamic of urban growth does not just influence land use changes but also
brings higher land prices for commercial use such as for malls replacing
residential use over time.
Many scholars of urban economics have studied on land market and
land values. The theories of land prices previously were taken from Ricardo
and Von Thunen in 19th century. These theories become a foundation of urban
land use and land values until now. Koomen and Buurman (2002) claim
Ricardian land models explain the existence of land rents from differences in
fertility or land quality. Land of a higher quality generates surpluses over land
with a lower quality. These surpluses are paid as rent to the landlord due to
competition in the land market and the market for agricultural products. In
addition, Von Thünen’s model is concerned with location and transportation
costs, which as well as fertility, are characteristics of a particular land parcel.
Though von Thünen merely analysed land use patterns, an important result of
his model was the explanation of land prices.
In addition, Ricardian and Von Thunen’s analysis according to Evans
(1983) were only based on the demand for land and the fixed supply is
1
assumed. That argument supports the static theory of land price. Moreover,
ibid (1983) suggest an alternative theory that incorporates supply side in the
analysis of land value. Thus, it will enrich the land value study.
Moving to the importance of land prices information, a tabulation of
urban land prices information is essential information for public or private
sector. For government, the latest and accurate land price information is
necessary to determine the expansion of infrastructure including the affordable
housing or transportation infrastructure. It is also important for taxation
system. In addition, private sector also uses the information for bank or
financial institution to assess the property values.
Some studies tried to collect information of land prices using different
approach such as market price or appraised price. However, Dowall and Leaf
(1991:707) indicated that the detailed information on how land prices vary
spatially within Third World cities was usually lacking. They added that no one
knows the spatial shape of land prices from the city centre to the suburb.
Furthermore, ibid (1991) argued there is still no adequate information how is
infrastructure provision and land title interact and influence land prices pattern
in growing Third World cities.
Noticing that, this study aims to examine the determinants and the
spatial pattern of urban land prices in one of the biggest cities in Indonesia that
is Surabaya. It is the second largest city in Indonesia after Jakarta, Surabaya
inhabits more than 3 million people. This city is located in northern shore of
eastern Java and accommodates the capital city of East Java. Regarding that, it
is worthy to study about land prices pattern and determinant variables of land
values in Surabaya. By knowing those, it will give more understanding about
the characteristics of urban structure in Surabaya. Spatial pattern of land prices
also can be used as an input for local government to impose zoning and land
market regulation.
1.2 Problem Statement
For more than a decade, Indonesia has succeeded improving its economic
growth and become one of the promising emerging countries. The vast area of
land which is approximately 1.9 million km 2 and has a population of more than
240 million people are one of the driving factors the economic from supply
and demand side. This leads to many investment opportunities in industrial or
service sector. Undoubtedly, this factor brings an expansion of business district
in large cities such as Jakarta and Surabaya, which are well known as
trading/service or business cities. As a result, rapid land use changes occur and
land prices are getting higher over time in Surabaya. However, transactions of
land are market based and not officially determined since there is still no detail
regulation related to land pricing in Indonesia, though there is zoning.
In addition, local newspaper Kompas, March 25 2011 reported that the
land price in Jakarta increased about 30% and generally 10% for entire
Indonesia. In Jakarta CBD that is Sudirman-Thamrin, the land prices constituted
2
almost 50 million per square meter. This phenomenon is not only in Jakarta
but also in metropolitan city such as Surabaya.
Based on Navastara and Navitas (2012), housing in Surabaya is quite
expensive because land availability is greatly decreasing. Consequently, higher
land prices would be a burden for Surabaya’s inhabitants to get affordable
housing. Building infrastructure such as transportation, parks and road could
be very expensive for the government as well. Moreover, increasing land prices
would be an obstacle for the provision of public services.
1.3 Research Question
1.3.1
Main Question
It is interesting to examine which factors determine the land values in
Surabaya. Some studies indicate that distances from CBD are still major factors
affecting the land values in some metropolitan areas such as New York,
Bangkok and Jakarta. By incorporating other factors, this research will trace
‘What are the factors determining the land prices in Surabaya?’. This question
will also help to answer type of urban structure of Surabaya whether it is
concentric, polycentric or multinuclei city.
1.3.2
Sub-Question
The sub-question aims to seek the spatial pattern of land values in Surabaya
which is ‘do land parcels in Central Surabaya have the higher value than in
suburban areas?’. By addressing this, it is expected that we can the
characteristic of the spatial pattern of urban land prices in Surabaya.
1.4 Structure of the paper
The beginning of this paper consists of an introduction including background,
problem statement and research questions. In chapter two, literature review
will be presented containing basic theory of land price, factor determined land
value from previous study and type of urban structure.
Chapter 3 shows the description of the study area in this paper which is
Surabaya. Methodology and data description will be followed in the next
chapter. It includes brief explanation of the data and the model. Then, results
and analysis will be given in chapter five. Lastly the concluding remark will be
drawn discussing whether Surabaya has different characteristics to other cities
globally and its implication for local land policies
3
Chapter 2
Literature Review
2.1 Basic theory of land value
To begin with, we would like to introduce the connection between land value
and rent. Theory of land price is closely related with rent particularly the rent
of agricultural land. The theory of urban land value is developed from the
theory of agricultural land pricing. Many scholars developed urban land price
model from Ricardian and Von Thunen’s work for agricultural land in 19 th
century. Based on Alonso (1974), the Ricardian theory of rent is based on the
fertility and productivity of land. Even Ricardo presented theory about
location, however, is focused on fertility of land. The theory of location then
developed by Von Thunen which assessed the differential of location on land
values. Over time, urban land price theory is largely developed from Von
Thunen’s study.
However, according to Evans (1983), as mentioned above the analysis
of Ricardian economist emphasized only on the demand of land and the supply
of land is ignored. Similarly, Von Thunen’s theory also suggest of demanddriven land prices. That theory can be called the static theory of the price of land.
Another theory is the dynamic theory that emphasizes differentiation between
price of land and the rent of land. This theory suggests that the prices of the
parcel of land depend upon the rents that may be obtained in the future years.
Ibid (1983) also argued that those theories could not explain various features
of land market because it is demand-oriented. He offered a new theory that is
the alternative theory which include supply side into the determination of land
price. By incorporating supply, the theory can explain that a tax on
development will decrease the rate at which land is developed and lead to an
increase in its price.
For Evans (1983), Von Thunen’s theory focused on differentiation of
the transportation cost to city centre for agricultural products. The rent that
farmers are willing to pay will decline with the distance. In other words, any
land very much further away from the city centre will not be cultivated and no
rent will be paid. Taking to urban realities, according to ibid (1983), the Central
Business District (CBD) is equivalent to the city centre in Von Thunen’s
model.
In addition, according to Koomen and Buurman (2002), the bid rent
theory is based on microeconomic theory. It was mainly developed in the
context of urban land uses and urban land values. Moreover, Fujita (1989)
stated that the bid rent function approach that was introduced into an
agricultural land use model by von Thunen (1826), and later extended to an
urban context by Alonso (1964) is essentially the same as the indirect utility
function approach, which was introduced into an urban land use model by
Solow.
4
In addition, the bid rent function explains the relation between urban
land uses and urban land values. In a very simplified view, households and
companies make a trade off between the land price, transportation costs and
the amount of land they use. Keeping in line with that, Cheshire and Sheppard
(1993) stated that urban land theory suggests that land value result from a trade
off between transport costs and accessibility. Similarly, Fujita (1989) also
explained, any household moving to the city choosing the residential location
will face a complex set of factors in making a decision. That is called trade-off
problem between three factors : accessibility, space and environment amenities.
Moreover, that household must also consider budget and time constraint.
2.2 Urban land price determinants and its analysis
Sufficient studies about urban land prices determination can be found taking
case in developed and developing countries. The urban land studies in
developed countries come from Brigham (1965), Peiser (1987) and Haughout,
et.al (2008). Some researches in metropolitan cities in urban developing
countries are conducted by, Dowall and Leaf (1991), Han and Basuki (2001)
and Wisaweisuan (2001). According to Han and Basuki (2001:1843), the
uneven distribution of land values has been attracted numerous studies in this
area to examine the factors influencing land values. They use different
approach and variables to examine factor influencing land prices. Below, we
have presented some studies related to land valuation determination.
Author/title
City
Variables
Methodology
Developing countries
Wisaweisuan (2001) /
Spatial characteristics of
land prices in Bangkok
Dowall and Leaf (1991)
/ The price of land for
housing in Jakarta
Han and Basuki (2001)
/ The spatial pattern of
land values in Jakarta
Bangkok
Lewis (2007) / Revisiting
the price of residential land
in Jakarta
Jakarta
Dowall (1992) / A
second look at the Bangkok
land and housing market
Bangkok
Jakarta
Jakarta
Proximity from CBD,
nearest sub centre and
infrastructure condition
Distance
to
CBD,
infrastructure provision
and land tenure
Distance
to
CBD,
nearest
highway,
commercial
establishment, zoning
and flood risk
Proximity from city
centre,
road
class,
environment condition
and land title
Distance to city centre
Non-linear regression
Multiple
analysis
regression
Stepwise
multivariate
regression model
Linear regression, nonlinear regression and
incidental
truncation
model
Non-linear regression
analysis
Developed countries
Haughout et.al (2008) /
The price of land in the New
York Metropolitan Area
New
York
Ahlfeldt and Wendland
(2008) / Fifty years of
Berlin
Type
of
property,
condition of property,
characteristics
of
transaction,
intended
use and transaction
Proximity from CBD
and travel time to CBD
5
Linear regression model
Non-linear regression
model
and
spatial
urban accessibility : the
impact of urban railway
network on the land gradient
in industrializing Berlin
Peiser (1987)
Brigham (1965) / The
determinants of residential
land values
autoregresive model
Dallas
metropoli
tan area
(nonresidentia
l)
Los
Angeles
county
Neighbourhood factors,
microlocation attributes,
development
expectations,
macrolocation,
macroeconomics
variables and physical
site characteristics
Accessibility, amenities
and topography
Linear
analysis
regression
Linear regression model
From all the studies presented above, the distance to city centre or
CBD is obviously the most popular variables in land valuation model.
According to Harvey (1992:223), “the CBD is the optimum location for shops,
commerce and services for, as the focus of inter-city transport, it has the
greatest accessibility. Competition for the limited space produces peak land
values and intensity of development in the form of skyscrapers or multi-storey
buildings”.
Distance to city center has become a dominant factor influencing land
prices in one city. There are some examples from developed countries related
to that. Brigham (1965) found that the price falls about four or five cents per
square foot per every one mile increase in distance from the CBD. In line with
that, Case and Mayer (1996) studied housing prices dynamic within
metropolitan area found that the distance to Boston which is the city centre
was also associated with the pattern of price changes. The results indicate that
house prices increased at a greater rate, the closer a town was to Boston.
Recent study from Houghwout, et.al in 2008 found that vacant land prices in
the New York Metropolitan Area were seen to decline with the distance from
the CBD. Nonetheless, the findings can be different for non-residential area.
Peiser (1987) which conducted a studies of non-residential land values in
Dallas found that distance to the CBD was only significant determinant for
office land but not for industrial land.
Similar findings related to the importance of city centre are also found
in developing countries such as the studies from Dowall and Leaf (1991), Han
and Basuki (2001), Lewis (2007) in Jakarta and Wisaweisuan (2001) in
Bangkok. They found that the distance from Central Business District (CBD)
is a dominant factor to explain land value variation. Along with these findings
from Jakarta, the satellite city of Jakarta that is Bekasi also experience the
similar relationship. Yowaldi (2012) found that the prices have negative
relationship with land price, when the distance to CBD increase the land value
declines.
In addition, Wisaweisuan (2001) explain that the importance of city
centre in land price variation indicate that a city supports the model of
monocentric assumption or core-dominated city. Mostly the results of these
6
papers reflect the neoclassic assumption of land price that indicate the positive
association between land price and the distance from city centre. Furthermore,
Ibid (2001) wrote that in neoclassical economics, market mechanism allocates
pieces of land to users on the basis of the bidding process. In other words, the
person who gives maximum bidding will always win the competition and be
able to get the piece of land for his or her intended use.
Another interesting feature from some studies is related to the change
of land price gradient in urban area. Land price gradient indicate the
importance of CBD to land prices. In the early of 20th-century in Chicago,
based on Yeates and Garner (1971) quoted from Han and Basuki (2001:1843),
distance variable (CBD) alone explain 75% of land values variation. Then in
1960, the distance could only expain 10% of the land price. Similar result can
be seen from Berlin, Ahlfeldt and Wendland (2008) using dataset from 1890 to
1936 found that a declining of land price gradient from a 77.5% in 1890 to
39.7% in 1936. These findings indicate the land gradient became flatter and
there is an increase in land values variation. Declining of the land gradient
from city centre is most likely due to the improvement of transportation
modes and agglomeration economy.
Declining importance of CBD over time can also be found in cities in
developing countries. For instance, in Jakarta, Lewis (2007:2191) reported that
the influence of the proximity to city centre has apparently become less
important overtime. Previous study from Dowall and Leaf (1991) showed that
62% of land price variation can be solely determined by the distance from city
centre. Ten years later, Han and Basuki found that the effect of CBD decline.
Ibid (2001) found only 30-45% of land price variation can be explained by
distance from CBD. The development of Jakarta has been reduced the spatial
linked between land price and city centre. It implies that there could be an
expansion of business districts in the periphery area of Jakarta.
The reduction of land prices gradient in some cities above indicates the
change of urban land structure from monocentric to polycentric city. This
transformation is most likely influenced by the improvement of transportation
network. Berlin and Istanbul experienced the transformation into polycentric
city. Based on Ahlfeldt and Wendland (2010), Berlin followed monocentric
pattern in 1890 and by 1936 the city structure became more complex and
resembling present-day polycentric cities. Istanbul also follow the similar
pattern with Berlin. According to Ciraci and Kundak (2000), “Istanbul has
been changing its monocentric structure to the polycentric structure”. Ibid
(2000) also explained the advantages of turning into polycentric in Istanbul
which are there is a deduction of transportation charge between the residential
areas and the CBD, reduce the consumption of fuel and the environmental
pollution.
It could be said that the flatter of land price gradient moves in the same
direction with the phenomenon of flattening out which reflects the impact of
increasing mobility and the decentralisation of urban activities. According to
Dowall (1992) comparing land prices gradient in Jakarta, Bangkok and Karachi
found that those cities have experienced a pattern of flattening out. Bangkok’s
7
land prices gradient is flattening out as suburban land prices increase faster
than those in city centre. Based on Dowall and Leaf (1991), Jakarta’s gradient is
far steeper than those recently estimated for Bangkok and Karachi are. These
differences are because of transportation accessibility and the patterns of the
density found in the cities. The analysis about the declining importance of
CBD and flattening out are very interesting to incorporate in Surabaya’s land
price analysis. However, due to the dataset in this research only one period, we
cannot include that kind of analysis.
Even the data does not allow us to analyze the effect of flattening out
or the declining of CBD effect on land prices variation to suburb area, our sub
research question tries to examine the differentiation of land value variation
between central and non-central part of Surabaya. We adopt this question from
Han and Basuki’s paper in 2001, where one of the purposes is to examine the
spatial pattern of land prices in central and non-central of Jakarta. By using
non-parametric tools that are Kruskal-Wallis and Mann-Whitney test, they
found that that land prices were not uniformly distributed among central and
non-central districts in Jakarta. In line with that finding, Navastara and Navitas
(2012:1) which examine the impact of residential land development in Surabaya
found that “the dynamics of land prices in Surabaya is quite disproportionate”.
By using dataset from May 2010 and September 2011, ibid (2012) found that
land values are only concentrated in three major assistant areas which are East
Surabaya, South Surabaya and West Surabaya.
Besides the distance from city centre or CBD, some studies try to
incorporate other distance and non-distance variables that might affect land
prices variation in their analysis. Distance or spatial variables consist of the
distance from nearest commercial buildings, openness area and nearest
highway. Non-distance variable including microlocation, macrolocation, travel
time to city centre, development expectation and macroeconomic variables.
Public transportation is closely linked with accessibility for resident
travelling to the city centre and other places. Ahlfeldt and Wendland (2008)
examine the travel time gradient found that there was a slight changes of travel
time gradient over time in Berlin. There is a connection between this result to
the flattening of land gradient. It indicates that transportation innovation
promote the the attractiveness of outlying area. There is an establishment of
business district in the peripheral area and hence the flattenning out of land
gradient instead of travel time gradient.
Other variable related to accessibility is the proximity to the nearest
highway. Han and Basuki (2001) found the negative sign of the nearest
highway to land values in Jakarta. It implies the greater the distance from the
nearest highway, the lower the land prices are. It is related to the unreliable
public transportation in Jakarta; hence, the residents choose a house that is
close to a highway.
In term of zoning, the findings from Haughwaout, et.al. (2008)
showed that residential property was found to sell for a slightly higher price
than commercial property; industrial land commanded a significantly lower
price than land designated commercial or residential. Keeping in line with that
8
result, Han and Basuki (2001) also found that a kelurahan (the lowest level
governmental region) in Jakarta with over 50% of its land for commercial
zoning had the higher level of land values than the one with other utilization.
In addition, street size does matter based on Peiser’s study (1987) in
Dallas Metropolitan Area. He found that street size was significant for both
industrial land and office land. Industrial land with frontage on a major arterial
is 43% higher, and that with expressway frontage is 68% higher, in value than
industrial land on a minor street. The premium for office land on major
arterials was about 38% higher than that on minor streets. Ibid (1987) also
examined macroeconomic variables in his research. Using dataset of
unemployment rate, change in unemployment, interest rate and change in
interest rate from 1978-1982, he found that commercial land value is more
likely to be influencing by macroeconomic condition compared to office and
industrial land. It is more likely that commercial land tend to influencing
quickly by worsening macroeconomic condition. Examining the role of
macroeconomic variable to land prices is worthy to study; however, it is likely
need time series dataset. Thus, this research does not include this variable.
Other determinants of land values, which found in some journals are
the infrastructure provision and land tenure. It can be found in journal from
Dowall and Leaf (1991) while taking case of Jakarta. The effect of
infrastructure was significant, which increases almost 50% to land prices. In
addition, they found that parcels of land with registered title, which can be
right-of-ownership or right-of-building, are 45% more valuable than those with
weak claim are. Furthermore, land title has a role intervening how the
infrastructure affect land values. It is because of clear title grants more secure
tenure and can be used for collateral in order to get bank credit.
Related to pollution, based on Ridker and Henning (1967) quoted from
Smith and Huang (1993) found that there is negative relationship between
property values and air pollution. Another factor that linked with land price
determination is openness area. Paper written by Shultz and King in 2001 has
grouped the openness area into 5 categories. They found that the signs on the
coefficients indicate that proximity to the large protected natural areas, golf
courses, and Class II wildlife habitats, as well as the percentage of vacant and
commercial land use, positively influences housing values. In contrast,
proximity to undeveloped and neighbourhood parks, Class I wildlife habitats,
as well as percentages of industrial land use and high levels of housing density,
negatively influences housing values. Regarding to the environment
neighbourhood, Mahan, et.al (2000) found that the size of the nearest wetland
increases house values.
One thing that is interesting to note in land valuation discussed above
is related to the methodology. Based on the list of literatures mentioned earlier,
there are two models commonly used in land values study. One of them is
non-linear regression that can be found in Dowall (1992). Other assumes that
land prices have linear relationship with distance, and then use the
straightforward linear regression model such as Houghwaut et.al (2008). In
addition, Lewis (2007) taking Jakarta as a case use both linear and non-linear
9
regression in his model and found that land price gradient in both model is not
remarkably different.
Another model that recently examined the relationship between land
prices and other variables particularly distance variables is spatial autoregressive
model. Ahlfeldt and Wendland (2008) traced out the land and travel time
gradient by comparing two models which are non-linear regression and spatial
auto regressive (SAR) model. SAR is used to reduce spatial dependency among
distance variable.
2.3 Type of urban structure
Type of urban land use pattern is closely related to land prices. Stated in
Harvey (1992) that ‘land uses determine land values and not vice versa’.
Therefore, knowing the pattern of land use is essential to understand land
value in this research. The general pattern of land use according to ibid
(1992:222) is the reflection of the demand and supply of the sites. Below, we
explain briefly 6 urban structure models which are :
1. Concentric zone theory or monocentric
According to Harvey (1992), this theory emerged from the examination of
the historical development of Chicago in the 1890s by E.W. Burgess (1925).
In essence, this model explains urban social within urban area. Figure 2.1
show the concentric zone.
Figure 2.1 Concentric urban zone
Source : Bitesite BBC
Four broad zones were postulated according to Bitesite BBC : (1) The
CBD; (2) The inner city, containing working-class population living in this
area with few amenities such as park ; (3) the inner suburbs, as
transportation modes improved and became cheaper the suburban areas
emerged containing newer and more spacious house; (4) the outer suburbs,
identified by good housing, amenities and environment. According to
Harvey (1992), this theory is lack of detail and fails to include various aspect
of special accessibility.
2. The radial or axial development theory
10
Figure 2.2 The radial urban zone
Source : Harvey (1992)
According to Harvey (1992), this theory generally modifies the concentric
zone model by include topographical features. Figure 2.2 show the radial
development that for example including the river which could distort
concentric zone as showed in figure 2.2 (a). The existence of the river will
cause the road and rail constructions to adjust it, and hence the zone to
assume a starfish shape. In summary, whilst concentric zone based on the
proximity in term of distance, axial theory is based upon accessibility in
terms of time, etc. (Harvey, 1992)
3. Cost of friction hypothesis
R.M. Haig (1926) developed this theory based on the Von Thunen model.
Harvey (1992) explained that this theory examined land values related to
transportation cost saved in different sites. The interesting thing from this
theory is that in any urban area competition will tend to produce the land
use pattern that minimises total sites rents and transport cost for that area
as a whole. (Harvey, 1992)
4. The wedge or radial sector theory
The radial sector model was the famous work of H.Hoyt (1939) and it is the
extending model of concentric model by allowing the development model
of a more irregular pattern.
11
Figure 2.3 Hoyt sector model
Source : Bitesite BBC
Figure 2.3 show the land use pattern is segregated by the social housing
class. Based on Harvey (1992), residential sector radiate outwards the CBD
constraining manufacturing or wholesaling into other sector. Therefore,
residential areas are segregated by income.
5. Multiple-nuclei theory
C.D. Harris and E.L. Ullman in 1945 in the article “The nature of the cities”
developed this ecological model. The theory still departed from concentric
zone model but takes the view that large cities have a structure which is
essentially nuclear.
Figure 2.4 Multi-nuclei urban model
Source : Harvey (1992)
The figure 2.4 above shows the multi-nuclei model where there is a
tendency of agglomeration. Some nuclei emerged as a minor settlement
12
before city growth or developing. The growth of population supports a
suburban shopping, business centre and industrial area. In addition, high
rents in the CBD force firms to move to the periphery area and the reliable
accessibility also lead the households and firms to concentrate in a particular
location. (Harvey, 1992)
6. Sector-zone theory
P.H. Mann (1968) developed this model taking British city as a case but this
theory considerably originated from a USA city. It is a combination of
concentric and sector zone and emphasizing on the social structure of the
area. Based on Harvey (1992), the locations of high-income residence is in
the west of the city and while the low-income to the east and near the
industrial sector. The figure 2.5 presents the sector-zone model.
Figure 2.5 Sector zone model
Source : Harvey (1992)
13
Chapter 3
Background of Study Area
Surabaya was officially established in 1293 and is the capital city of East Java
province. Surabaya is a coastal city that is located on the northern shore of
eastern Java along the edge of the Madura Strait. The regional border of this
city can be seen below :
North : Madura Strait
East : Madura Strait
South : Sidoarjo Regency
West : Gresik Regency
This city forms parts of a greater metropolitan area which is called
Gerbangkertosusilo including Lamongan to the northwest, Gresik to the west,
Bangkalan to the northeast, Sidoarjo to the South and Mojokerto and Jombang
to the southwest. In 2008, Greater Surabaya Metropolitan Area or
Gerbangkertosusilo has an area of around 10.332 km2 and the population is
11.625.512 people. It is counted 50% contribute to the regional income in East
Java Province. Map 3.1 below present the region of Gerbangkertosusilo.
Map 3.1 Great Metropolitan Surabaya Area
Source : National Land Agency
14
From all of the Gerbangkertosusilo regions, Surabaya has a relatively high
economic growth compared to the others. Moreover, it contributed to more
than 26% to regional income of East Java in 2008. Surabaya obviously has an
importance role in East Java’s economic as a whole. It has been a connector
between west and east region of Indonesia through Tanjung Perak Port since a
long time ago. In addition, economic condition in this region is slightly better
compared to others. Based on Statistics Indonesia, the economic growth in
Surabaya in 2011 was 7.52% and that number was higher compared to East
Java and national growth. In addition, the economic condition mostly depends
on trading and service. The structure of its economy was driven by tertiary
sector that accounts to 60.61% of Gross Domestic Regional Product (GDRP)
in 2011.
Geographically, this city has 5 major assistant areas which are North,
South, West, East and Central Surabaya, 31 districts and 163 Sub-districts. The
administrative map of Surabaya is presented below.
Map 3.2 Administrative Map of Surabaya
Source : Author’s illustration based on land values and land use maps
Based on municipality’s data, the number of Surabaya inhabitants in
June 2013 is 3.163.645 person. It keeps increasing overtime as the population
growth between 2000 and 2010 is 0.63% per annum. Population Density in
this region 7.966 people per km2, 8,642 people per km2 in 2000 and 2010
respectively, and the most dense area in Surabaya is in South Surabaya.
The characteristic of physical land in Surabaya is mostly flat land with
the inclination less than 3% and 20% of the Surabaya’s area have low wave
with inclination about 5-15%. In addition, land use can be grouped into two
15
major factors that are agriculture and non-agriculture. The latter utilization
accounts 30.076,30 ha from the total area that is 36.508,39 ha. Further,
according to Sari (2011) the land cover for residential area has increased over
time which 4.907 ha in 2002 to 5.595 ha in 2009.
16
Chapter 4
Data and Methodology
4.1 Data
In this research, we use secondary spatial and textual data. Constructing spatial
data is quite challenging. Firstable, employing Google Earth and other sources
of information in order to determine the exact location of independent
variables are. Then, Arc GIS (mapping software) is applied to process land
value map, zoning map, land use map and spatial images. It is also used to
measure the distance between all samples to predictor points and layout all of
the maps. Lastly, converting the spatial data into textual data by using textual
software.
Below, detail information is given related to the data :
1.
Land value map
This map is obtained from National Land Agency’s land valuation survey
in 2011. Land values in this survey are based on market price. The
procedure of land valuation survey can be seen as follow :
 Delineating the land price zone based on the utilization of land using
aerial photo mapping or satellite image, land use map, zoning map,
registration land map and administration map. In every zone, the land
price are expected relatively the same or not diverse.
 To collect the market land price in every zone then samples are
determined. Samples are the parcel of land that give the information
of transaction and bidding price in the last 24 months for nonagriculture land and 48 months for agriculture land. The transaction
and bidding price can be obtained from the local government officer.
 Samples are chosen by stratified random sampling technique. For
every zone, at least 3 samples are taken. The vacant land is preferable
to choose as a sample. The samples should be fairly distribute in every
zone and should be fit with the utilization/zoning of land.
In Surabaya’s survey, they took 392 observations. All of the
observations we take as our samples; hence, the unit of observation in this
research is parcels of land. This dataset have a weakness of the accuracy of
land use. As stated above that they are taken the samples based on the
existing land use but there are some mistakes in plotting it. For example, it
is clearly from satellite image that the existing land use is industrial area
but they wrote as residential zone. Map 4.1 below show the output of
Surabaya’s land valuation survey in 2011.
17
Map 4.1 The land valuation map
Source : National Land Agency’s land valuation survey
The red area indicates the highest price and the dark green is the
lowest. The other colours indicate the middle range of prices. In addition,
the west shape (mangrove forest) is omitted in this map because this
survey differ an ordinary region and a particular region such as forest.
They have not surveyed and counted for mangrove forest value yet.
2. Satellite images
This data is necessary to plot the independent variables such as public
facilities, openness area, port, and other images. It is also used to examine
the accuracy of other maps such as land value or land use map. Satellite
images are taken from Google Earth 2013.
3. Land use map.
Based on regulation, land use is the land cover of the earth. This map is
surveyed in 2009 and divided the area into 7 existing utilizations of land in
Surabaya that are industry, residential, agriculture (rice field) and
agriculture (annual crop), farm, inland water and open area. However, only
four categories of land use are recorded the samples in this study which
are industry, residential and agriculture for rice field and annual crop.
18
Map 4.2 Land utilization of Surabaya, 2009
Source : National Land Agency (BPN)
Map 4.2 indicate that more than two third of the existing land use is
for residential area which is shown by dark pink. The other utilizations is
shown by bright pink (industrial area), bright blue (inland water) and green
(agricultural). Some utilization cannot be seen clearly since the scale of the
map and the area of those utilizations is small.
4. Zoning map indicates the expected zoning/spatial planning for this city
for 5 years. This map is supposed to revise in 2012 but until now the
arrangement have not imposed yet in local parliament. It is created in 2007
consisting eight zones, which are industrial area, green open area,
residential area, trading/services area, public utilities, port-planning area,
particular area (military zone, port) and conservation area (lake). The
samples are located in seven of eight zoning designation except for
extended port-planning zone.
19
Map 4.3 Zoning Map of Surabaya
Source : National Land Agency (BPN)
As stated in Lewis (2007) that local authorities plan urban land uses
poorly in Jakarta. That situation is not different with Surabaya’s condition.
Zoning regulation is often ignored and violated both by the residents and
by the government. For instance, green open space based on regulation is
supposed about 30% from total area in a region. However, according to
municipality data, the green area in 2011 is approximately 20.25%. Even
though the data indicate there is still lack of green area, there is a
significant progress on it. In 1995, total green area was accounted only
3.172 Ha or 9.6% of Surabaya’s area.
5. Land title or land tenure
Quoted from Dowall and Leaf (1991), property rights are weakly defined
in Indonesia. There are three types of land title for resident in Indonesia
which are right of ownership (hak milik/HM), right to build (hak guna
bangunan/HGB) and right to use (hak pakai). Those titles are registered by
BPN (Badan Pertanahan Nasional) or National Land Agency. However,
some parcels are still not registered. Therefore, we designated land tenure
into two categories which are registered and not registered. In this dataset,
there are two registered titles recorded that are right-of-ownership and
right-to-build types of land tenure.
We would explain briefly about HGB and HM. Right of ownership is
hereditary right, the strongest and the most complete right over land that
can be posed only by Indonesian citizen. Right of ownership can be
removed if the land falls to the government due to revocation for public
purposes, voluntary acquisition, abandoned land, and foreigner ownership.
20
HGB means that the holder can establish buildings on land, which is
not his/her, own within period of 30 years. By considering the concerned
parties’ request and the condition of buildings, the right of building can be
extended for another 20 years. This right can be removed due to
expiration period, inappropriate conditions with the terms, released by the
holder, revocation for public purposes, abandoned land, and foreigner
ownership.
6. Road Class
Road variables are closely linked with the accessibility of residence to
travel. In this study, road class consist of 4 classes depends on which part
of parcel located in the type of road. The classes of road are more
specifically explained below :
1 =
footpath (2-3m)
2 =
local road (2-6m)
3 =
collector road (4-12m)
4 =
artery road (8-12m)
Footpath is the road that only can be used by pedestrian; Local Street is
the street that is primary served vehicle with speed low and short route;
collector road is the public street that serves traffic from local road to
arterial road; artery road is the high capacity urban road.
7. CBD (Central Business District) and sub CBDs
Surabaya, as the second largest city in Indonesia, experienced fast
development in trading sector. Therefore, the municipality of Surabaya
identified seven Central Business Districts which are :
1. The Kembang Jepun CBD, was the eldest trade region in the
Surabaya city
2. Tunjungan – Embong Malang - Basuki Rahmad CBD
3. Ngagel CBD, was the relocation from the industry region
4. Kertajaya CBD
5. Jemursari CBD
6. Mulyosari CBD
7. Mayjen Soengkono – HR Mohammad CBD
Of seven CBDs, we appointed Tunjungan as the centre of Surabaya’s
CBD. The reason lies of the choice because there are shopping centres
and hotels in that area. Furthermore, there is Tunjungan Plaza which is the
new icon and the biggest shopping centre in Surabaya. The others we
chose as sub-CBDs. For Mayjen Soengkono and HR Mohammad, we
divided into 2 sub-CBDs. Therefore, there are seven sub CBDs in the
map. A property’s distance from the central business district is measured
from the property in question to a point in the middle of the central
business district.
21
Map 4.4 Location of CBD and sub-CBDs
Source : Author’s illustration based on land values and land use maps
8. Distance from traditional market
As stated above that economic of Surabaya is driven by trading and
service sector. Hence, we include trading places or market as one of our
variables. Recently, many malls and plazas established in Surabaya but
traditional market still exist and contribute to the economic. Traditional
markets that we chose are the markets with the highest transaction. They
are pasar atum, pasar turi, pasar keputran, pasar pabean and pasar genteng. The
location of those markets is quite near to city centre as we can see in map
4.5 below.
22
Map 4.5 Location of traditional market
Source : Author’s illustration based on land values and land use maps
9. Distance from public transportation infrastructure
The near distance from public transportation indicates the accessibility of
people to travel to other places. Surabaya also has complete public land,
sea and air transportations that could serve the local, regional and
international trip. One thing that is worthy to note that structure
economic of Surabaya is depended upon industry and trade sector. Then,
transportation infrastructure that carrying the products outside Surabaya is
imperative. Thus, I chose four modes transportation that serve people and
products to other regions of Surabaya which are port, bus and train station
and one essential bridge which is Suramadu bridge or Jembatan Suramadu.
There is one port in Surabaya which is Pelabuhan Tanjung Perak built
in 19th century. It is one of gateways of Indonesia, which becomes
collector and distribution of goods from and to eastern Indonesia.
International Juanda Airport is located in Sidoarjo that is the satellite city
of Surabaya and part of Surabaya Greater Metropolitan Area. Even
located in Sidoarjo, this airport mainly served people who are going to
Surabaya.
Only two bus stations that we include in this research considering
those are the major bus station type A. They are Terminal Purabaya and
Terminal Tambak Oso Winangun, which serve local and regional routes.
However, Purabaya, the biggest bus station in Surabaya, is not located in
Surabaya. It is located in Sidoarjo, precisely near the south border between
those cities.
Similar with bus station, we only counts two train stations that are
Stasiun Gubeng and Stasiun Pasar Turi. Gubeng serve the routes going to
23
south parts of Java whereas Pasar Turi serves the trains that are going
from Surabaya to the north regions of Java.
Jembatan Suramadu is the first bridge that connecting two islands in
Indonesia which are Java and Madura. Suramadu stands for Surabaya
Madura Bridge linking Surabaya in the east of Java to Bangkalan, Madura.
It is essential to stimulate the economic growth in Madura. Not only as
connector, this bridge is also known as the new icon tourism of Surabaya.
Map 4.6 Location of public transportation and infrastructure
Source : Author’s illustration based on land values and land use maps
10. Distance from industrial area (pollution indicator)
Two industrial zones in Surabaya are Rungkut and Tambak langon-KalianakMargamulya.
24
Map 4.7 Location of industrial area
Source : Author’s illustration based on land values and land use maps
11. Distance from rubbish dump
Along with the increase of population in Surabaya, the volume of bargage
also increases over time. Furthermore, location of rubbish dump site
become a problem. Until October 2011, open dumping land (TPA) was
located in Sukolilo district but local people parry all of garbage. Nowadays,
Surabaya has one active location of open dumping land, which is called
TPA Benowo. It is located in the fringe of this city that is in Benowo
district, West Surabaya. The large area is approximately 26 Ha and it can
accommodate waste more less 2,390 tons per day.
25
Map 4.8 Location of final disposal site
Source : Author’s illustration based on land values and land use maps
12. Distance from openness area
Openness area is related with amenities. Openness area in this research
consists of public parks, golf courses and undeveloped city forest,
mangrove forest. There are 9 city parks including in this study and those
are mainly located in city centre. In addition, 4 private golf arenas are
existing in Surabaya. In fact, there are two undeveloped city forest, which
are Balas Klumprik and Pakal. Due to the location of Pakal is not clear
enough in municipality’s website so that we only include Balas Klumprik in
this study.
26
Map 4.9 Location of openness area
Source : Author’s illustration based on land values and land use maps
4.2 Methodology
In analysing the spatial pattern of land value distribution and determinant in
Surabaya, firstly, we will digitise the points (samples) in land value map and
then divided into 5 sub-regions (districts). The attributes in every point then
are analyzed by applying geographic information systems (GIS) software. GIS
mapping functions, together with descriptive statistics, are used to plot and
analyse the spatial patterns of the various variables such as distance from CBD
and other variables.
Regarding model in land valuation studies, Ahfeldt and Wendland
(2008) suggest that land value might be well described by an exponential
function of the distance to CBD. The exponential functional relationship is
usually estimated using the well-established log-linear specification :
LV is land value; α corresponds to the log of land value in the city center; β is
the percentage change in land value as one move 1 km away from the CBD
and
is an usual error term.
Lewis (2007) also use log-linear specification for basic land prices
model that can be written as follow :
27
where, Pi is the price of land i; Ai is parcel size;
are the distance of land to
the city centre;
are additional determinants of the unit price of land; are
the associated coefficients to be estimated; and is the error term. This model
has been used of many studies including for Jakarta, which is Dowall and Leaf
in 1991.
Therefore, in this study we use natural logarithm transformations to
address the main question. The multiple regression models is applied to
evaluate the relationships between the land values with various independent
variables. The model can be written as follow :
In addition, natural logarithm transformations for land price and distance
variables aiming for convenient means of transforming extreme value
differentiation in land values and the distance.
Table 4.1 Dependent and independent variables
Variables
Detail
Dependent variable
Price/square foot(V)
Price per square foot in Rupiah
Independent Variables
 Distance from CBD (C)
Direct linear distance from CBD
 Distance from traditional market
(M)
Direct linear distance from traditional
market
 Distance from sub CBD (SC)
Direct linear distance from sub CBD
 Location of road (R)
Dummy (Er1)=1) for footpath
Dummy (Er2)=1) for artery
Dummy (Er3)=1) for collector road
road
 Land Use (LU)
Dummy (Elu2=1) for agriculture (field
rice)
Dummy (Elu3=1) for industry
Dummy (Elu4=1) for residential
 Zoning (Z)
Dummy (Ez1=1) for conservation area
(lake)
Dummy (Ez2=1) for green open area
Dummy (Ez3=1) for industrial area
28
Dummy (Ez4=1) for particular area
(military zone, port)
Dummy (Ez5=1) for public utilities,
Dummy (Ez6=1) for residential area
 Pollution (P)
Direct linear distance from industry
area
 Land title (L)
Dummy variable denoting for nonregistered (0) and registered (1)
 Distance from public
transportation
Direct linear distance from port (PO)
Direct linear distance from bus
station (BS)
Direct linear distance from train
station (TS)
Direct linear distance from airport
(AP)
Direct linear distance from Suramadu
Bridge (SB)
 Distance from rubbish dump
Direct linear distance from rubbish
dump/final disposal site (R)
 Distance from openness area
Direct linear distance from parks(PK)
Direct
linear
distance
from
undeveloped city forest (UF)
Direct linear distance from golf
courses (GC)
Direct linear distance from mangrove
forest (MF)
For sub-question, we will examine the question using the Mann-Whitney and
Kruskal-Wallis test to confirm whether land value in central district differ from
suburban area.
4.3 Data description
As stated above that the number of samples is 392 observations spreading in 5
major assistant area and 31 districts. Table 4.2 show the summary statistics of
the data. It is clearly seen that the land prices are highly distant between the
minimum and maximum. Mean value of land price in this research is Rp
2,334,994. It is similar with the finding from Navastara and Navitas (2012)
which evaluate land prices dynamic in Surabaya. Ibid (2012) found that overall
land prices are dominated by 2 to 3 million rupiah per square meter.
29
Table 4.2 Summary statistics of land values in Surabaya
Variable
Obs
Mean
Std. Dev.
Min
Max
Dependent
Price/m2 (Rp)
392
2,334,994
1,585
431,288
8,721,607
Independent
Distance from CBD (m)
392
6,051
3,355
266
15,072
Distance from sub CBD (m)
392
3,002
2,210
13
11,638
Distance from traditional market
(m)
392
5,047
3,483
46
14,153
Pollution
Distance from industrial
area (m)
392
5,681
2,328
831
11,230
Public Infrastructure and
Transportation
Distance from train
station (m)
392
5,430
3,370
298
13,761
Distance from bus station
(m)
392
7,929
2,851
1,546
14,816
Distance from port (m)
392
10,371
3,991
918
18,172
Distance from airport (m)
392
14,969
4,536
5,377
27,438
Distance from Suramadu
Bridge (m)
392
10,182
4,546
599
19,240
Distance from rubbish dump
392
13,598
4,571
2,163
23,821
Openness Area
Distance from park (m)
392
4,340
3,205
152
15,446
Distance from golf course
(m)
392
5,702
2,842
353
11,499
Distance from mangrove
forest (m)
392
11,227
4,984
1,639
24,936
3,540
225
16,147
Distance from
undeveloped city
392
9,040
forest (m)
Source : Author’s elaboration based on land values data
Moreover, the near distance for some independent variables are quite
diversed. It can be noticed in proximity to Juanda Airport, Jembatan Suramadu,
TPA Benowo and mangrove forest. It is due to the location of those variables
are in the fringe of Surabaya Greater Metropolitan Area, hence the near
distance from samples are quite diverse.
30
To get better understanding about the distribution of land prices, map
below is presented to see the range of prices. We divided the samples into 5
group which the highest number samples is in red (more than Rp 5 million)
and the lowest is in the bright yellow (less than Rp 1 million). It also indicates
that the higher land value concentrate in the central and the lower is in the
outlying area of Surabaya. Moreover, the higher land prices in Surabaya move
from central region to north, south and east part of this city.
Map 4.10 Distribution of land prices
Source : Author’s illustration based on land values and land use maps
Table below presented detail information related the distribution of
land prices among the samples.
Table 4.3 Distribution of land prices
Percent
Category
Obs
10.46
< Rp 1,000,000
41
43.62
Rp 1,000,000-2,000,000
171
25.51
>Rp 2,000,000-3,000,000
100
11.73
>Rp 3,000,000-5,000,000
46
8.67
>Rp 5,000,000
34
Source : Author’s elaboration based on land values data
Average prices (Rp)
758,587
1,484,821
2,337,644
3,842,016
6,465,114
Of 5 categories, more than half or almost 70% of the samples are
mostly distributed in the mid range group which are between Rp 1 million to
Rp 3 million. Only less than 20% of the samples are accounted in the lowest
and highest level of land prices.
31
To know the differences of land value in central and suburban area in
Surabaya, the summary statistics is presented below.
Table 4.4 Summary statistic of land value in central and non central area
Non-Central
Central
Overall
North
West
South
East
Minimum
(n = 35)
961,527
(n = 357)
574,256
(n=64)
611,317
(n = 87)
431,288
(n = 89)
583,779
(n = 117)
670,638
Maximum
8,674,385
6,166,464
8,378,282
2,876,298
8,721,607
7,364,722
Range
7,712,858
5,592,209
7,766,965
2,445,010
8,137,828
6,694,084
Mean
3,709,336
2,148,794
2,665,849
1,408,694
2,165,936
2,560,272
Standard
Deviation
2,532,620
1,169,213
1,887,377
592,559
1,212,153
1,390,254
Source : Author’s elaboration based on land values data
Table 4.4 shows that land values in the central area constitute the
highest mean value that is around Rp 3.7 million per square meter. The overall
land prices of non central is approximately 2,1 million rupiah per square meter.
Furthermore, the non-central region that is West Surabaya reports the lowest
prices. In essence, there are a substantial different between central and noncentral land prices.
Another interesting thing from table 4.4 is that one of the land parcels
in South Surabaya recorded the highest prices. In line with that, Navastara and
Navitas (2012) found that South Surabaya accounted highest level of land
prices growth compared to other regions. It indicates that there are growing
middle class residential in South Surabaya.
32
Chapter 5
Result and Analysis
5.1 Descriptive Analysis
Before examining the result by regression analysis, we will describe the data
with Exploratory Data Analysis (EDA). This tool helps to analyze the main
characteristics of the data with visual methods such as graphs and tables.
5.1.1
CBD and land price
City centre has become the major features in urban land prices studies. Based
on some studies, the CBD usually has important role in determining land
prices. Before using regression, we will examine the prices distribution to city
centre based on the samples characteristic that can be seen in table 5.1. Then,
to get further understanding, we use spatial figure by creating four rings of land
prices surrounding the CBD.
Table 5.1 Average land prices based on the distance to the CBD
Distance (km)
Freq
Percent
Average Prices
0-5
172
43.88
2,841,990.57
5.1-10
164
41.84
2,128,500.49
10.1-15
56
14.29
1,382,520.57
Source : Author’s elaboration based on land values data
Table 5.1 show that land prices tend to decline with distances to city
centre. The higher distances from CBD, the prices are getting lower. This
finding in line with other studies from Peiser (1987), Dowall and Leaf (1991),
Han and Basuki (2001), and Haughwaout (2008) which also found that the
land values declined with the distance from city centre.
Related to the pattern of the samples and CBD, we created land prices
ring map. There are four rings that in every ring have different large of area.
The first ring has the smallest area and the outer ring/the fourth ring has the
largest area. This map mainly adopted from the ring zones of the Paris city,
which the area to the outer is getting larger. Further analysis will be given
related to the outliers in our samples.
33
Map 5.1 Ring of land prices distribution to CBD
Source : Author’s illustration based on land values data
It is clearly seen that in the first ring the land prices are more likely vary
than other circles. The pattern of the land prices in map tend to gradually
decline with the proximity to city centre. Parcels of land that are located in
CBD show the highest prices indicated by the red point. However, there is one
sample in yellow one that is very unlikely expected to locate there.
Table 5.2 shows the mean, minimum and maximal values of samples in
every circle. There are five parcels that are exactly located in CBD area
(Tunjungan – Embong Malang - Basuki Rahmad). Furthermore, the samples are
mostly located in third and fourth ring that is located in the fringe area of
Surabaya.
Table 5.2 Summary statistic of land prices’ ring
Ring
Freq.
Mean
Min
CBD
5
5,859,939
961,527
1
23
4,583,458
1,639,369
2
94
2,821,256
670,638
3
170
2,101,295
513,308
4
100
1,581,802
431,288
Source : Author’s elaboration based on land values data
Max
8,078,195
8,721,607
8,378,282
6,558,450
5,536,754
By noticing table 5.2 And map 5.1, it can be said that in CBD there is
one parcel that has extremely low value compared the other four samples. The
land price of that sample is less than one million rupiah per square meter. In
contrast, the mean value of the samples located in CBD is almost 6 million
rupiah. In addition, there are five samples that the land prices are substantially
different from the other parcels in the first ring.
34
Table 5.3 Detail information of outliers
No
Ring
Prices (Rp)
Land Tenure
Land use
Road class
City
1
CBD
961527
HGB
Residential
Local
Central Surabaya
2
2
819902
HGB
Residential
Local
East Surabaya
3
2
670638
HGB
Residential
Local
East Surabaya
4
2
934268
HM
Residential
Local
East Surabaya
5
2
940734
HM
Residential
local
North Surabaya
6
2
863580
HGB
Residential
local
North Surabaya
Source : Author’s elaboration based on land values data
After consulting with satellite images of outliers using Google Earth
2013 which can be seen in Appendix 1, we are hardly to determine what factor
cause those land plots have outlier prices. It is difficult to predict without
conducting the real survey why those observations were distant from the rest
of the data. Nonetheless, there are some predictions that we would like to
explain related to that :
1. Conflicting land
There would be a case that the landowner does not control or abandoned
the land for many years. Even though there is a regulation that the owner
should control its land, some landowners violate that. As a result, other
people illegally occupy and control that parcels. That would be a problem
for the owner to sell the land due to the complicated land controlling
problem.
2. Land title life span and zoning regulation
The land tenure of the outliers are right-of-ownership (HM) and right-ofbuilding (HGB). We only examine relation of HGB due to this right has a
life span. HGB title is granted by the National Land Agency for an initial
period of up to 30 years and is extendable for a subsequent 20-year period.
Perhaps, the HGB title for parcels above is almost expired and need to readminister. There would be a cost to extent it. Another thing that likely
related to is the interaction of land title and zoning regulation. It is
possible that local government plans to change the city spatial planning in
that area. That causes the uncertainty for HGB right’s holder and they sell
the land for cheap prices.
3. Physical factors
Physical factors such as located in the block that prone to flooding, slum
area and prone-fire could be lower the land prices. As suggested in Han
and Basuki (2001) that a kelurahan with flood risk are likely has lower value
compared to a kelurahan without that risk.In addition, the lower land prices
could be influenced by the existence of home industry in the
neighbourhood. Since small home industry such as Batik also likely
produces pollution.
Another physical factor that could be affected is the parcels are located
near military base, intelligent or secret agency’s office. It is possible based
35
on some experiences of people living near Indonesia intelligent office in
Jakarta that they experience bad signal of communication. There would be
some services such as internet or phone that are likely disturbed by located
near that kind of office.
5.1.2
Non-distance variables
There are four non-distance variables in this study, which are land use, zoning,
road type and land title. Land use and zoning have some similarities since in
arranging zoning policy; existing land utilization is one of the inputs.
Table 5.4 Average land price based on land use and zoning classification
Category
Freq
%
Average Price/m2 (Rp)
Land Use
Agriculture, Annual Crop
Agriculture, Rice Field
Industry
Residential
8
7
14
363
2.04
1.79
3.57
92.6
1,543,553
1,134,726
1,780,731
2,396,958
2 0.51
18 4.59
15 3.83
10 2.55
15 3.83
273 69.64
59 15.05
1,488,029
1,790,148
1,694,799
3,131,690
2,795,569
2,169,642
3,205,664
Zoning
Conservation Area
Green open space
Industrial Area
Particular Area (Military, Port)
Public facilities
Residential Area
Trading/Services
Source : Author’s elaboration based on land values, land use and zoning data
Industrial area in land use classification is consisted of agricultural
industry, trading and service, whereas for zoning map, the data differentiate the
industrial area and trading/service. It is because the nature of dataset of land
use obtained from the entire of East Java Province. Meanwhile, zoning map is
obtained for Surabaya city. Therefore, zoning map is more detail in classifying
than land use.
There are two major features that interesting to explain from the table
above which are residential, industry and trading/service area. Land prices for
residential area in both land use and zoning utilization record relatively higher
prices compared to other classifications. Furthermore, for land use
classification, it is accounted as the highest land price. As it was suggested in
literature from Haughwaout, et.al (2008) that residential plots in New York
were sold with higher prices compared to other categories.
For zoning, the highest values are the area designated for
trading/services (commercial land). This finding is similar with the condition in
Jakarta based on Han and Basuki’s paper in 2001. Ibid (2001) found that area
which design for commercial zone was likely have the higher value. In other
words, both cities commanded higher land prices for commercial zoning than
other classifications.
36
Table 5.5 show the distribution of land prices relative to the road class
and land title. Of the 4 road categories, most of the samples are located near or
in the front of local road. In addition, the parcels that located near or in front
of artery road recorded the highest prices. It is likely that road type does matter
to land prices in Surabaya. That finding is in line with the paper from Peiser
(1987) indicating that non-residential land plots in major arterial are higher
than in the minor street.
Table 5.5 Average land prices related to road class and land tenure
Road Class
Land Title
Average
Average
Category
Freq
Price/m2
Category
Freq
Price/m2
(Rp)
(Rp)
Footpath
17
1,949,406
Registered
117
2,468,307
 HGB
Local
286
2,250,026
271
2,279,527
 HM
Collector
81
2,627,329
Artery
8
3,232,066
Non
Registered
4
2,193,479
Source : Author’s elaboration based on land values data
Table 5.5 above indicate that mostly samples are already registered or
have official land title. There are two types of registered land tenure in the
samples, which are right to build (HGB) and right of ownership (HM). Only
four samples have not registered yet. Moreover, parcels that have registered
land title commanded slightly higher prices compared to non-registered
parcels. Parcels with registered land title in Jakarta according to Dowall and
Leaf’s journal published in 1991 are 45% more valuable than those with nonregistered. From explanation above, Surabaya has the same pattern related to
land tenure that registered land title has the higher prices that non-registered
parcels.
In addition, land plots that registered as HGB get higher values than
HM. The drawback of land plots with HGB is that the holders should readminister before the span of the certificate end. However, apparently the
Surabaya’s resident does not take into account HGB or HM. In addition, many
real estate developers nowadays sell a house with HGB title since the state
cannot grant the right of ownership to private sector. However, the buyers as
individual can convert the certificate from HGB to HM in the next few years.
5.1.3
Distance variables
We identified 13 distance variables except proximity to CBD in this analysis
that are related to public transportation and infrastructure, openness area and
final disposal site (rubbish dump). We divided the variables into two categories
which are the prices tend to increase and decrease with distance. It is necessary
to distinguish the variables in order to see the different feature of land prices
between them.
37
Distance to port, train station, Suramadu Bridge, industrial areas, parks,
markets, mangrove forest and golf courses are likely to decline with distance.
However, for mangrove, Suramadu Bridge, industrial region and golf courses
in the middle range (5.1-10 km) account the highest prices. It we look back to
the map in chapter 4, those variables are located in the suburb area of Surabaya
but the proximity to city centre is in medium distance, which around 7 to 10
km (table 5.6). In other words, there is likely intercorrelation of land prices
between the distances of these variables to city centre. Below, we present the
proximity of 13 variables to city centre
Table 5.6 Average distance from variables to CBD
Variables
Distance (m)
Undeveloped city forest
9,330.47
TPA Benowo
12,980.04
Bus station
16,196.54
Parks
2,402.15
Train Station
1,058.77
Port
7,145.75
Market
1,744.42
Golf courses
7,997.11
Industrial Area
7,325.17
Mangrove
9,949.37
Airport
15,404.23
Sub CBDs
4,481.05
Suramadu
6,977.65
Source : Author’s elaboration based on land values data
Table 5.6 indicates that some variables are located far away from CBD
signed by normal case. Variables that close to CBD which in range (0-5 km)
signed by italic case and variables that the proximity in the middle range signed
with bold case. Further illustration distance variables that the prices falls with
distance can be seen in figure 5.1.
Figure 5.1 Land prices that decline with distance
Golf
Price/m2
Industry
Price/m2
3,000,000
3,000,000
2,500,000
2,500,000
2,000,000
2,000,000
1,500,000
1,500,000
1,000,000
1,000,000
500,000
500,000
0
0
0-5
5.1-10
10.1-15
Distance (Km)
0-5
5.1-10
Distance (Km)
38
10.1-15
Suramadu
Price/m2
Sub CBD
Price/m2
3,500,000
3,000,000
3,000,000
2,500,000
2,500,000
2,000,000
2,000,000
1,500,000
1,500,000
1,000,000
1,000,000
500,000
500,000
0
0
0-5
5.1-10
10.1-15
0-5
Distance (Km)
Port
Price/m2
3,500,000
5.1-10
10.1-15
Distance (Km)
Market
Price/m2
3,000,000
3,000,000
2,500,000
2,500,000
2,000,000
2,000,000
1,500,000
1,500,000
1,000,000
1,000,000
500,000
500,000
0
0-5
5.1-10
0
10.1-15
0-5
Distance (Km)
Train
Price/m2
5.1-10
10.1-15
Distance (Km)
Park
Price/m2
3,000,000
3,000,000
2,500,000
2,500,000
2,000,000
2,000,000
1,500,000
1,500,000
1,000,000
1,000,000
500,000
500,000
0
0
0-5
5.1-10
0-5
10.1-15
Distance (Km)
5.1-10
Distance (Km)
Mangrove
Price/m2
3,000,000
2,500,000
2,000,000
1,500,000
1,000,000
500,000
0
0-5
5.1-10
Distance (Km)
Source : Author’s elaboration based on land values data
39
10.1-15
10.1-15
Land prices related to proximity to port, train stations, markets, parks
and Sub CBDs fall gradually with distances. The reason behind that is these
variables are located near the CBD except port that located in the middle range
(7 km) if we consult the map in chapter 4 and table 5.6. As a result, the bar
chart for Tanjung Perak Port in figure 5.1 is quite steeper compared to train
stations, parks and sub CBDs.
The opposite direction with figure 5.1, the bar charts below present
four distance variables that land prices are likely to increase with distance.
Figure 5.2 Land prices that rise with distances
Price/m2
Undeveloped forest
Bus Station
Price/m2
3,000,000
3,000,000
2,500,000
2,500,000
2,000,000
2,000,000
1,500,000
1,500,000
1,000,000
1,000,000
500,000
500,000
0
0
0-5
5.1-10
0-5
10.1-15
Distance (Km)
Airport
Price/m2
5.1-10
10.1-15
Distance (Km)
Rubbish Dump
Price/m2
3,000,000
3,000,000
2,500,000
2,500,000
2,000,000
2,000,000
1,500,000
1,500,000
1,000,000
1,000,000
500,000
500,000
0
0
0-5
5.1-10
10.1-15
Distance (Km)
0-5
5.1-10
Distance (Km)
Source : Author’s elaboration based on land values data
Figure 5.2 show that undeveloped forest, bus stations, rubbish dump
and airport are the variables that tend to increase the land prices with distance.
Location factor could be a reason why these variables tend to increase with
distance. Furthermore, the interaction of those variables and the proximity to
city centre could be one of the factord of the increasing trend of land prices
with distance. Location of bus stations, one of the bus stations that are Terminal
Purbaya, is not in Surabaya but in Sidoarjo regency (satellite city of Surabaya)
and Terminal Tambak Oso Winangun is in suburban area of Surabaya. In addition,
the location of airport is in Sidoarjo and quite far from Surabaya so that there
are no samples are located 0-5 km from airport.
40
10.1-15
For undeveloped forest, the location is in South Surabaya and quite far
to the city centre. Rubbish dumpsite is also located in periphery area that is
West Surabaya, which is the region that accounted the lowest level of land
prices compared to other regions in Surabaya. Moreover, the land parcels near
final disposal site tend to increase substantially with distance. It is likely due to
environmental reason since people are unlikely to choose and avoid living in
the sites near open dumping area.
5.1.4
Central and non-central analysis
Examining the spatial pattern of central and non-central in Surabaya, we use
non-parametric Mann-Whitney and Kruskal-Wallis test.
Table 5.7 Result of Mann-Whitney test
Obs
Rank sum
Non-central
357
68152
Central
35
8876
-3.124***
z
0.0018
Prob > z
***indicates significant at the 1 percent level
Expected
70150.5
6877.5
This z-value confirms that there is a statistically significant difference
between the distributions of the land prices of central and non-central region.
After comparing the rank sum and the expected rank sum, the rank of land
prices for central area is higher compared to non-central area. Therefore, the
central group has the higher prices than non-central area
Table 5.8 Result of Kruskall-Wallis test
City
Obs
West Surabaya
87
Central Surabaya
35
South Surabaya
89
East Surabaya
117
North Surabaya
64
chi-squared
61.7
with 4 d.f
Probability
0.0001***
***indicates significant at the 1 percent level
Ranksum
10323.5
8876.0
17796.5
27147.0
12885.0
The results from Kruskal-Wallis test indicate to reject null hypothesis
that is the samples of all groups come from identical population. It confirms
that the chi-squared with four degree of freedom is significantly different from
zero. It implies that land prices in 5 major assistant area are not equally
distributed. This finding is also supported from Navastara and Navitas (2012)
which found that the outlying region in Surabaya that is South Surabaya, West
Surabaya and East Surabaya dominated the land prices growth in the period
2010 to 2011. It implies that there is uneven growth of land prices in Surabaya.
Furthermore, ibid (2012) also found that South Surabaya experienced the
41
highest growth suggesting that development of new housing in that region is
more massive than other regions.
Along with that, land values among central and non-central districts in
Jakarta also were not uniform in Jakarta based on Han and Basuki (2001). They
found that at 1 percent level, the prices in central districts were statistically
higher than non-central districts. Furthermore, the land values were not
uniformly distributed among non-central districts.
5.2 Regression result and analysis
In this part, the regression result and its discussion will be presented as follow.
5.2.1
Regression of land prices and CBD
This partial regression only accounts one independent variable that is
the CBD. The reason is that some studies put this variable as the main analysis.
In addition, it is also interesting to see the relationship between land values and
the CBD in Surabaya and compared to the results with other cities such as
Jakarta. This regression use robust standard error to remove heteroskedasticity
problem.
Table 5.9 Regression result of land prices and CBD
Variable
Coefficient
Constant
17.452
ln dist CBD
-0.349
2
R
0.1716
***indicates significant at the 1 percent level
t-stat
45.41***
-7.87***
This regression result indicate that coefficient of CBD in Surabaya is
negative and significant suggesting that land values decline with the distance
from city centre. The coefficient of CBD is -0.349 implying that 10% increase
of proximity to city centre is likely decline the land prices, on average, by 3.5%.
Similarly, taking example from Jakarta in Dowall and Leaf (1991) paper show
that land values and distance from CBD is negative and highly statistically
significant.
In addition, the R-squared for Surabaya is 0.17 implying approximately
17% of land prices variation can be explained by the distance from city centre.
If we compared to other findings, this value is quite small. Dowall and Leaf
(1991) indicated that 62% of variation in land values in Jakarta can be solely
explained by the distance from CBD. Furthermore, there was a reduction of
the variation related to CBD in Jakarta based on Han and Basuki study in 2001.
They found there is a declining of the importance of city centre in Jakarta.
Moreover, ibid (2001) confirmed only 30-45% of land prices variation can be
determined by the proximity from city centre. In other words, land price
gradient become flatter over time in Jakarta. Flattening of land price gradient
also found in Berlin based on research from Ahlfeldt and Wendland (2008).
Nonetheless, the finding from Jakarta and Berlin are still higher than in
Surabaya.
42
Wisaweisuan (2001) wrote that the importance of city centre to land
prices indicate that the urban structure of one city support the model of
monocentric assumption or core dominated city. Considering the finding from
Surabaya, the R-squared is quite small compared to other city. It indicates that
this city does not follow core dominated city structure. Moreover, it could be
said that the pattern of urban structure in Surabaya followed multi-nuclei city
or polycentric assumption. However, we cannot say that Surabaya has
transformed to polycentric city such as Istanbul (Ciraci and Kundak (2000))
and Berlin (Ahlfeldt and Wendland (2010)). Since, we do not have any spatial
history data of land prices there.
In addition, the lower of R-squared above indicated there is a
decentralisation of urban activities in Surabaya. This argument is supported by
Navastara and Navitas (2012), which study land prices dynamics in Surabaya.
They found that residential development moves to suburb area of Surabaya
which are East Surabaya, South Surabaya and West Surabaya. Furthermore, if
we notice in chapter 4, there are seven sub business districts spreading in
Surabaya.
To know the other variable that influence land prices variation in
Surabaya, we incorporate other variables in the next regression.
5.2.2
Pooled regression
The relation between dependent and all independent variables can be
seen in the result of pooled regression below. It shows that CBD, the
significant variable in land prices study, is not statistically significant. The
variables that significant are distance from sub-CBD, bus station, Suramadu
Bridge, industrial area and open dumping site, zoning (industrial area), and
road (collector).
Table 5.10 Pooled regression result in land value determination
Variables
Coefficient
t-stat
ln dist CBD
-0.146
-0.94
ln dist sub CBD
-0.119
-2.62***
Transportation
 ln dist train
-0.076
-0.71
 ln dist bus
-0.294
-2.34**
 ln dist port
0.018
0.09
 ln dist suramadu
0.342
3.34***
 ln dist airport
0.235
0.88
Openness
 ln dist parks
0.062
1.05
 ln dist golf
0.057
0.62
 ln dist mangrove
-0.206
-1.49
 ln dist undeveloped forest
0.049
0.59
ln dist market
-0.144
-1.36
ln dist rubbish dump
0.511
2.79***
Pollution
43
 ln dist industrial
Road
 R1 (footpath)
 R2 (artery)
 R3 (collector)
Land Title
 lt1 (non-registered)
Zoning
 Z1 (conservation)
 Z2 (green open area)
 Z3 (industry)
 Z4 (particular area)
 Z5 (public utilities)
 Z6 (residential)
Land use
 lu2 (agricultural, field rice)
 lu3 (industry)
 lu4 (residential)
Constant
R-squared
***indicates significant at the 1 percent level;
5 percent level.
0.177
2.08**
-0.075
0.194
0.164
-0.63
1.05
2.25**
0.042
0.47
-0.344
-0.156
-0.379
0.144
-0.033
-0.113
-1.67
-1.13
-2.42**
0.87
-0.23
-1.28
-0.080
-0.38
0.415
1.88
-0.028
-0.18
9.787
1.81
0.367
**indicates significant at the
After examining the regression above, we found that the result violates
the OLS assumptions of multicollinearity. It implies that multiple regression
model contain high correlation between independent variables; hence, the
coefficient regression is biased. To avoid those, principal component analysis
(PCA) can be applied to reduce some variables from equation. Appendix 2
indicates the result of multicollinearity test using VIF (variance inflation
factor). A VIF of 5 or 10 above indicates high multicollinearity. The result
shows that 10 of 14 distance variables constituted high score of VIF that are
more than 5.
5.2.3
Principal component analysis (PCA)
The aim of PCA is to find uncorrelated variable; hence avoid
multicollinearity issue. In essence, PCA ensure the linear combinations chose
have maximal variance. We conduct PCA only for distance variables that are
most likely correlated with each other. There are three steps in doing PCA
which can be seen below.
Step 1 : Examine the eigenvalues to determine how many principle
components can be retained.
Table 5.11 Eigenvalues, and the proportion of variation explained by the
principal components.
Component
Eigenvalue
Proportion
Comp1
5.628
0.402
44
Comp2
Comp3
Comp4
Comp5
Comp6
Comp7
Comp8
Comp9
Comp10
Comp11
Comp12
Comp13
Comp14
2.990
2.076
1.127
0.586
0.545
0.342
0.239
0.120
0.114
0.086
0.074
0.045
0.029
0.214
0.148
0.081
0.042
0.039
0.024
0.017
0.009
0.008
0.006
0.005
0.003
0.002
From the proportion of Eigenvalues, the first four components account more
than 80% of variance. Therefore, four components should be considered.
Step 2 : After determining four principal components, running PCA with four
components. Next, we can look at the coefficients for the principal
components. In this case, since the data are standardized, within a column the
relative magnitude of those coefficients can be directly assessed.
Table 5.12 The principal component coefficients
Variable
Comp1 Comp2
Ln dist CBD
0.466
-0.033
Ln dist market
0.456
-0.060
Ln dist train
0.441
-0.047
Ln dist park
0.432
0.187
Ln dist Sub CBD
0.360
0.125
Ln dist rubbish
-0.058
-0.553
Ln dist mangrove
-0.042
0.538
Ln dist airport
-0.036
0.511
Ln dist golf
0.096
-0.135
Ln dist undeveloped city forest
0.045
0.030
Ln dist suramadu
0.113
0.047
Ln dist port
0.237
-0.233
Ln dist industry
0.082
0.009
Ln dist bus
-0.040
-0.138
Comp3
0.091
0.002
-0.035
0.201
0.044
-0.068
-0.243
-0.001
0.631
0.625
-0.271
-0.252
-0.233
0.109
Comp4
-0.004
0.041
0.069
0.076
0.123
0.190
0.037
0.217
-0.165
-0.172
-0.217
-0.002
0.805
0.481
Values greater than 0.4 have been bold
Interpretation the principal components based on which variables are
most strongly correlated with each component (the numbers that are large in
every column). The first component are corelated with some of the observed
variables. The remaining components that are extracted in the analysis display
the same two characteristics: each component accounts for a maximal amount
45
of variance in the observed variables that was not accounted for by the
preceding components, and is uncorrelated with all of the preceding
components.
Due to distance from Sub-CBDs, Suramadu Bridge and port are not
correlated with other variables, and then we exclude those variables, and then
run the PCA from step 1. Table 5.13 show the principle component coefficient
with 11 variables
Table 5.13 the principle component coefficient with 11 variables
Variable
Comp1 Comp2 Comp3 Comp4
Ln dist CBD
0.5254 -0.0368 0.0221 -0.0009
Ln dist market
0.5047 -0.0655 -0.0574 0.0197
Ln dist train
0.4994 -0.0591 -0.1011 0.0480
Ln dist park
0.4520 0.1984 0.1717 0.1039
ln dist rubbish
-0.0311 -0.5944 -0.1275 0.1796
Ln dist airport
-0.0739 0.5362 0.0831 0.2401
Ln dist mangrove
-0.0400 0.5302 -0.2126 0.0184
Ln dist undeveloped city forest
-0.0315 0.1060 0.6807 -0.1004
Ln dist golf
0.0255 -0.0828 0.6285 -0.1091
Ln dist industry
0.0628 0.0000 -0.1548 0.7970
Ln dist bus
-0.0824 -0.1214 0.1822 0.5178
Values greater than 0.4 have been bold
Step 3 : Component Summaries
 First Principal Component Analysis
The first principal component is a measure of the distance of market, train
and CBD, and parks. If we look back to the map in chapter 4, the location
between those variables is quite near to each other. All variables are
positively related as the signs are positive.
 Second Principal Component Analysis
The second principal component is a measure of the distance of airport,
mangrove and rubbish. All of the signs are positive implying those
variables positively related. Furthermore, consulting the map in the
previous chapter, we suspect the distance all of the samples to these
variables are likely not very different. If we see the mean distance from
table 4.2, it can be seen that the mean value of those variables are not
distant.
 Third Principal Component Analysis
The variables of the openness area that are proximity to golf courses and
undeveloped city forest are correlated in third PCA.
 Fourth Principal Component Analysis
The fourth principal component is a measure of the proximity of
industrial area and bus station.
5.2.4
Regression using principal components
From the result of the PCA, we would group the principle components above
based on the mean distance that can be shown in table 4.2. The first PCA is
46
grouped as center area, the second PCA as the outer area and the third and
fourth as the middle area 1 and 2. Below regression result is presented by
incorporating principal component and non-distance variables.
Table 5.14 Regression result using principle components
Variables
Coefficient
t-stat
Component 1 (central area)
-0.115
-7.140***
Component 2 (outer area)
-0.110
-6.640***
Component 3 (middle area 1)
-0.030
-1.450
Component 4 (middle area2)
-0.006
-0.270
Road
 R1 (footpath)
-0.022
-0.190
 R2 (artery)
0.258
1.370
 R3 (collector)
0.173
2.440**
Land Title
 Lt1 (non-registered)
0.080
0.770
Zoning
 Z1 (conservation)
-0.132
-0.910
 Z2 (green open area)
-0.161
-1.170
 Z3 (industry)
-0.351
-2.420**
 Z4 (particular area)
0.189
1.140
 Z5 (public utilities)
-0.078
-0.520
 Z6 (residential)
-0.138
-1.550
Land use
 Lu2 (agriculture, field rice)
-0.069
-0.350
 Lu3 (industry)
0.273
1.250
 Lu4 (residential)
-0.051
-0.320
Constant
14.595
80.750***
R-squared
0.305
***indicates significant at the 1 percent level; **indicates significant at the 5
percent level
There are four independent variables that are significant from table 5.6
which are principal component 1 (center area), principal component 2 (outer
area), zoning (industrial area) and road class (collector). R-squared is 0.305
suggesting that this model can explain 30.5% of land prices variation.
Distance from first component is negative and statistically different
from zero. The coefficient of this variable suggests that any 10% increase in
distance from centre area, the land values falls about 1.12%. This result fits
with theory and other empirical studies from Brigham (1965), Case and Mayer
(1965), Haughwaut et.al (2008), Dowall and Leaf (1991), Han and Basuki
(2001) and Wisaweisuan (2001). They found that any increase to the proximity
to city centre will likely decline the land prices in urban area.
The result from second component comes from three distance
variables that are rubbish dump, airport and mangrove. The second
47
component that are groupped as distance variables in outer area get
regression’s coefficient, which is negative and significant. It can be interpreted
that any increase of distance from outlying area will decrease land prices. The
coefficient is -0.110 that suggest that 10% increase of distance to outer area
likely fall the land prices, on average, 1.10%. This result is obviously
unexpected because we suspect that the farther away from outlying area, the
prices will increase.
The underlying argument that can be given for unexpected sign is that
if we see again the figure 5.2, the prices for mangrove variables tend to fall with
distance. However, the other variables (airport and open dumping site) tend to
increase with distance. The location of those variables are likely also influence
the result except for open dumping site since the variables located in the area
that is likely has lower value which are West Surabaya. Mangrove forest is
located in east of Surabaya and not too far from CBD. Airport is located in
southeast of Surabaya and the distribution of land prices in the middle range to
the outer range is not quite differ. From the analysis above, the mixed of these
variables in one component result in unreliable sign of the regression’s
coefficient.
Road also has an effect in land prices. The coefficient of R3 (collector)
is positive indicating that land prices of parcel located in front of collector road
is higher than located in local road. Based on Peiser (1967) which trace out the
determinants of land prices in non-residential area in Dallas found that road
type is significant for both industrial land and office land. Industrial land with
frontage on a major arterial is 43% higher, and that with expressway frontage is
68% higher, in value than industrial land on a minor street.
In addition, zoning for industrial area is significant and negative. It
implies that plot in industrial area has lower price compared to trading/service
land parcel. It is likely because trading/service zone in city spatial planning
map is located in city centre. Furthermore, the industrial zone is designed in
the fringe region of Surabaya. The finding from Jakarta also shows similar
result. As stated in Han and Basuki (2001), kelurahan that more than 50% of its
area is designated for commercial land use was likely having higher value
compared to other utilization.
48
Chapter VI
Conclusion and recommendation
6.1 Conclusion
As mentioned by Dowall and Leaf (1991) that “detailed information as to how
land prices vary spatially within Third World cities is usually lacking”. Hence,
this research try to examine the spatial and non-spatial factors influencing the
land prices pattern in one of the biggest in Indonesia which is Surabaya. This
study use market-based prices dataset to determine what factor that likely
influence land prices. It also aims to examine the spatial pattern of land prices
and in Surabaya.
Concerning the data and its interaction with CBD, it tends to some
observations move in the opposite direction. We identified six samples that the
land prices are extremely low or the values are very distant to others.
Addressing why those samples have the outliers prices from others without
conducting real survey is difficult. Nevertheless, some predictions are given
which are related to conflicting land, physical factor, land title and zoning
regulation.
We design 14 distance variables and 4 non-distance variables that likely
have an impact on land prices. Land prices for 10 distance variables tend to
decrease with distance and others are likely rise with distance based on
descriptive analysis. In term of non-distance variable, land use and zoning
category indicate that land parcels for residential purposes accounted relatively
higher prices. Nonetheless, land plots designated for trading/service
(commercial) get the highest prices. In addition, the parcels that have not
registered yet commanded lower prices compared to registered land. In
addition, road class also influence the land prices. The land plots in front of
collector road are appreciated with higher prices.
Even the urban growth is seemingly distributed to suburban area in
Surabaya, the result from Mann-Whitney and Kruskal Wallis test confirms that
land prices between central and non-central area of Surabaya are not equally
distribution. It is undoubtedly that there is still uneven distribution of land
prices. Thus, land parcels near city centre are still appreciated higher prices
than in non-central area.
Based on some studies in urban land prices, the distance from CBD
still significantly affects the land prices in the cities in both developed and
developing countries. Similar with other cities such as New York and Jakarta,
city centre has significant effect influencing land prices in Surabaya. The
regression result between land values and the CBD is negative and significant
implying that the increasing proximity from city centre, the lower the land
prices are. However, the R-squared for Surabaya is quite smaller compared to
its counterpart that is Jakarta. Only 17% of land prices can be solely explained
by the distance from city centre in Surabaya. Han and Basuki (2001) indicate
49
that 30-45% of land prices variation in Jakarta can be explained by CBD. This
finding suggests that rapid urban growth is likely distributed to outlying region
of Surabaya.
As stated that above that R2 in Surabaya is smaller than Jakarta, it is
likely that urban planning in Surabaya tend to adopt polycentric city. It can be
seen that local government in Surabaya appointed 7 business districts spreading
around the city. The emergence of sub business districts indicates there is a
decentralisation of urban economy or agglomeration economy. As a result
there is an increase in land prices variation which denoted by a lower Rsquared.
Incorporating all independent variables that are 14 distance variables
and 4 non-distance variables in regression analysis, result in multicollinearity
problem. Hence, principal component analysis (PCA) is applied to reduce
variables which are highly correlated. Based on PCA result, four principal
components can be included in the regression. Those principal components
can be grouped into central, middle and outer area. Conducting regression with
the principle components result in the significant of four independent variables
which are first principal component or centre area, second principal
component or outer area, zoning for industrial area and collector road.
It can be concluded that the regression result by incorporating principal
component fit with the findings from other journals except fo the principle
component 2. This component consists of three variables that are open
dumping site, mangrove forest and airport. Those variables have different
pattern of land prices. The parcels near mangrove forest tend to decline prices
whereas others are likely increase with distance. As the consequence, this
coefficient regression is unexpected. In addition, the R-squared in this model is
0.3337 implying that 33.37% of land prices variation can be explained by this
model.
Overall, the findings of this research related to distance to city centre
does not differ with other studies. CBD still influence the land prices variation
in Surabaya, similar with the findings from other city such as New York,
Dallas, Jakarta and Bangkok. However, the magnitude of CBD in Surabaya is
quite smaller compared to other cities. It indicates that this city does not follow
monocentric assumption. It is interesting to examine in the future land prices
study related to agglomeration economy and urban decentralisation in
Surabaya.
Regarding to future research in land prices determination; it is
interesting to examine the phenomenon of the declining importance of city
centre and the flattening out in urban activities over time. Furthermore,
examining spatial history of land use and transportation network can sharpen
the analysis of urban land prices. Instead of using direct linear distance from
transportation modes, travel time probably can be used in future study. In
addition, it is worthy to incorporate social variables such as crime and poverty
rate, social economic class in land prices study. It could be interesting to see
the interaction of social variables to dynamic of urban land prices.
Macroeconomic variables could be considered in future study. Since land as an
50
asset is closely related with credit bank and other macroeconomic variables. In
term of methodology, spatial econometrics could be applied to further study to
enrich the discussion and findings in urban land prices study.
6.2 Recommendation
Dowall and Leaf (1991) stated that land market fast-growing Third
World cities is terribly disorganised. It can be seen from land market in
Indonesia. The transaction of a parcel depends on the free market based on the
highest bidding. As a result, a gradually increase of land prices in urban area
cannot be controlled. It would be a problem in the future for government and
residents. Since higher land prices are likely a restriction in building public
utilities or infrastructure.
In addition, the pattern of land prices in Surabaya shows that this city
does not follow the pattern of monocentric city. Since there are an
establishment of sub CBDs around the city, transfer urban growth to outlying
area in Surabaya. Recently the concentration of residential growth based
Navastara and Navitas (2012) move to non-central area implying there is an
improvement in accessibility and shortened distance to urban area in Surabaya.
It also indicates of the improvement of public service provision in the
periphery area.
It is suggested that detailed zoning in the future to support the equal
development between central and non-central area in Surabaya. The provision
of green open spaces could be increased in this city. Moreover , the expansion
of infrastructure in the outlying of Surabaya could be accelerated to support
the economic activities. Hence, the uniformly development will likely avoid
sprawl pattern in this city.
51
Appendices
Appendix 1 : Satellite images of the outliers
Outlier 1
Green bullet indicate the parcel’s location
Outlier 2, 3 and 4
Green bullet indicate the parcel’s location
52
Outlier 5
Green bullet indicate the parcel’s location
Outlier 6
Green bullet indicate the parcel’s location
53
Appendix 2 VIF result
Variable
ln dist airport
ln dist rubbish
ln dist market
ln dist port
ln dist Suramadu
ln dist CBD
ln dist mangrove
in dist train
ln dist bus
ln dist park
in dist undev forest
ln dist golf
in dist industrial area
ln dist sub CBD
Lu3
Z3
Z6
Z2
R3
Z5
R1
Z1
Z4
Lu2
R2
Lu4
Tenure1
Mean VIF
VIF
19.58
16.74
14.38
7.13
6.09
13.12
9.64
8.46
7.02
4.89
5.36
5.07
3.65
2.75
4.08
2.27
2.09
1.55
1.37
1.3
1.27
1.25
1.26
1.96
1.1
3.81
1.07
6.28
54
References
Ahlfeldt, G. M. and Wendland, N. (2008) ‘Fifty Years of Urban Accessibility:
The Impact of Urban Railway Network on the Land Gradient in
Industrializing
Berlin’, University
of
Hamburg
Working
Papers from Chair for Economic Policy No 23. Hamburg.
Ahlfeldt, G. M. and Wendland, N. (2010) ‘How polycentric is a monocentric
city? The role of agglomeration economies’, Accessed 18 August 2013
<http://mpra.ub.uni-muenchen.de/id/eprint/24078>.
Ai, P. (2005) ‘Residential Land Value Modelling, Case Study of Hankou,
China’. Master thesis, International Institute for Geo-Information
Science and Earth Observation, Enschede.
Alonso, W. (1964) Location and Land Use. Cambridge : Harvard University
Press.
Case, K.E., and Mayer, C.J., (1996) ‘Housing price dynamics within a
metropolitan area’, Regional Science and Urban Economics, 26 : 387-407.
Ciraci, H. and Kundak, S. (2000)‘Changing Urban Pattern Of Istanbul; From
Monocentric To Polycentric Structure’ ERSA conference papers from
European Regional Science Association (August 2000).
Cheshire P., and Sheppard, S. (2005) ‘The Introduction of Price Signals into
Land Use Planning Decision-making: A Proposal’, Urban Studies 42:
647.
Dowall, D.E., and Leaf, M. (1991) ‘The Price of Land for Housing in
Jakarta’,Urban Studies, 28(5) : 707-722.
Dowall, D.E. (1992) ‘A Second Look at the Bangkok Land and Housing
Market’ Urban Studies 29(1) : 25-37.
Evans, A.W. (1983) ‘The Determination of the Price of Land’, Urban Studies 20
: 119-129.
Fujita, M. (1989) Urban Economic Theory Land use and city size. Cambridge :
Cambridge University Press.
Han, S.S. and Basuki, A., (2001), ‘The Spatial Pattern of Land Values in
Jakarta’, Urban Studies, 38(10) : 1841–1857.
Brigham, E.F., (1965) ‘The Determinants of Residential Land Values’, Land
Economics, 41(4) : 325-334.
55
Harvey, J. (1992) Urban Land Economics, Hampshire : MacMillan.
Haughwout, A., Orr, J., and Bedoll, D., (2008), ‘The Price of Land in the New
York Metropolitan Area’, Federal Reserve Bank Of New York, 14(3).
Koomen E. and Buurman J. (2002) ‘Economic Theory and Land Prices in
Land Use Modeling’, paper presented at the 5th AGILE Conference
on Geographic Information Science, Palma (Balearic Islands Spain)(
25-27 April).
KS3 Bitesite BBC (2013) ’Settlement in Urban Area’, accessed from
<http://www.bbc.co.uk/bitesize/ks3/geography/spaces/settlement_u
rban/revision/4/>
Laoli, N. (2011) ‘Harga Tanah di Jakarta Naik di Atas 30 Persen’, 25 March,
accessed from <http://entertainment.kompas.com/read/2011/03/25/
14105972/Harga.Tanah.di.Jakarta.Naik.di.Atas.30.Persen
Mahan, B.L., Polasky, S., and Adam, R.M., (2000) ‘Valuing Urban Wetlands: A
Property Price’, Land Economics, 76(1) : 100-113.
Navastara, A.M. and Navitas P. (2012) ‘Impact of Residential Land
Development towards Land Price Dynamics in Surabaya’, paper
presented at the 11th IRSA International Conference (July 2012).
Peiser, R.B., (1987) ‘The Determinants of Nonresidential Urban Land Values,
Urban Economics, 22 : 340-360.
Sari, L.I.P., (2011) ‘Evaluasi Perkembangan Tutupan Lahan Dan Pertumbuhan
Penduduk Menggunakan Regresi Linier di Surabaya Timur’, Master
Thesis, Institut Teknologi Sepuluh Nopember, Surabaya.
Shultz, S.D., and King, D.A., (2001) ‘The Use of Census Data for Hedonic
Price Estimates of Open-Space Amenities and Land Use’, Real Estate
Finance and Economics, 22:2/3 : 239-252.
Smith, V.K., and Huang, J.C. (1993) ‘ Hedonic Models and Air Pollution:
Twenty-Five Years and Counting’, Environmental and Resource Economics,
3: 381—394.
Wisaweisuan, N. (2001) ‘Spatial Characteristics of Land Prices in Bangkok’,
Thammasat Review, 6(1) : 142-176.
Yowaldi, (2012) ‘The Relation between Land Price and Distance to CBD in
Bekasi’, Master thesis, International Institute of Social Studies, The
Hague.
56